Help Center/ ModelArts/ ModelArts User Guide (Standard)/ Using ExeML for Zero-Code AI Development/ Using ExeML for Image Classification/ Creating an Image Classification Project
Updated on 2024-10-29 GMT+08:00
View PDF
Share

Creating an Image Classification Project

ModelArts ExeML supports sound classification, text classification, image classification, predictive analytics, and object detection projects. You can create any of them based on your needs. Perform the following operations to create an ExeML project.

Procedure

  1. Log in to the ModelArts console. In the navigation pane, choose ExeML.
  2. Click Create Project in the box of your desired project. The page for creating an ExeML project is displayed.
  3. On the project creation page, set parameters by referring to Table 1.
    Table 1 Parameters

    Parameter

    Description

    Name

    Name of an ExeML project

    • Enter a maximum of 64 characters. Only digits, letters, underscores (_), and hyphens (-) are allowed. This parameter is mandatory.
    • Start with a letter.
    • The name must be unique.

    Description

    Brief description of a project

    Dataset

    You can select a dataset or click Create Dataset to create one.

    • Existing dataset: Select a dataset from the drop-down list box. Only datasets of the same type are displayed.
    • Creating a dataset: Click Create Dataset to create a dataset. For details, see Creating a Dataset.

    Output Path

    Select an OBS path for storing ExeML data.

    NOTE:

    The output path stores all data generated in the ExeML project.

    Training Flavor
    Select a training flavor for this ExeML project. You will be billed based on different flavors.
    NOTE:
    • You can choose the package that you have bought. In the configuration fee area, you will see your remaining package quota and how much you will pay for any extra usage.
  4. Click Create Project. Then, the ExeML workflow is displayed.
  5. Wait until the workflow of the image classification project executes the following phases in sequence:
    1. Label Data: Check data labeling.
    2. Publish Dataset Version: Publish a version for the labeled dataset.
    3. Check Data: Check whether any exception occurs in your dataset.
    4. Classify Images: Train the dataset of the published version to generate a model.
    5. Register Model: Register the trained model with model management.
    6. Deploy Service: Deploy the generated model as a real-time service.

Quickly Searching for a Project

On the ExeML overview page, you can use the search box to quickly search for and filter workflows based on the ExeML type (or project name).

  1. Log in to the ModelArts console. In the navigation pane, choose ExeML.
  2. In the search box above the ExeML project list, filter the desired workflows based on the required property, such as name, status, project type, current phase, and tag.
  3. To adjust the basic settings of ExeML and select the columns you want to see, click on the right of the search box.

    Table Text Wrapping: This function is disabled by default. If you enable this function, excess text will move down to the next line; otherwise, the text will be truncated.

    Operation Column: This function is disabled by default. If you enable this function, the Operation column is always fixed at the rightmost position of the table.

    Custom Columns: By default, all items are selected. You can select columns you want to see.

    Figure 1 Customizing table columns
  4. Click OK. Then, the columns will be displayed based on the settings.
  5. To arrange ExeML projects by a specific property, click in the table header.